|
""" |
|
ein notation: |
|
b - batch |
|
n - sequence |
|
nt - text sequence |
|
nw - raw wave length |
|
d - dimension |
|
""" |
|
|
|
from __future__ import annotations |
|
|
|
import torch |
|
from torch import nn |
|
|
|
from einops import repeat |
|
|
|
from x_transformers.x_transformers import RotaryEmbedding |
|
|
|
from model.modules import ( |
|
TimestepEmbedding, |
|
ConvPositionEmbedding, |
|
MMDiTBlock, |
|
AdaLayerNormZero_Final, |
|
precompute_freqs_cis, get_pos_embed_indices, |
|
) |
|
|
|
|
|
|
|
|
|
class TextEmbedding(nn.Module): |
|
def __init__(self, out_dim, text_num_embeds): |
|
super().__init__() |
|
self.text_embed = nn.Embedding(text_num_embeds + 1, out_dim) |
|
|
|
self.precompute_max_pos = 1024 |
|
self.register_buffer("freqs_cis", precompute_freqs_cis(out_dim, self.precompute_max_pos), persistent=False) |
|
|
|
def forward(self, text: int['b nt'], drop_text = False) -> int['b nt d']: |
|
text = text + 1 |
|
if drop_text: |
|
text = torch.zeros_like(text) |
|
text = self.text_embed(text) |
|
|
|
|
|
batch_start = torch.zeros((text.shape[0],), dtype=torch.long) |
|
batch_text_len = text.shape[1] |
|
pos_idx = get_pos_embed_indices(batch_start, batch_text_len, max_pos=self.precompute_max_pos) |
|
text_pos_embed = self.freqs_cis[pos_idx] |
|
|
|
text = text + text_pos_embed |
|
|
|
return text |
|
|
|
|
|
|
|
|
|
class AudioEmbedding(nn.Module): |
|
def __init__(self, in_dim, out_dim): |
|
super().__init__() |
|
self.linear = nn.Linear(2 * in_dim, out_dim) |
|
self.conv_pos_embed = ConvPositionEmbedding(out_dim) |
|
|
|
def forward(self, x: float['b n d'], cond: float['b n d'], drop_audio_cond = False): |
|
if drop_audio_cond: |
|
cond = torch.zeros_like(cond) |
|
x = torch.cat((x, cond), dim = -1) |
|
x = self.linear(x) |
|
x = self.conv_pos_embed(x) + x |
|
return x |
|
|
|
|
|
|
|
|
|
class MMDiT(nn.Module): |
|
def __init__(self, *, |
|
dim, depth = 8, heads = 8, dim_head = 64, dropout = 0.1, ff_mult = 4, |
|
text_num_embeds = 256, mel_dim = 100, |
|
): |
|
super().__init__() |
|
|
|
self.time_embed = TimestepEmbedding(dim) |
|
self.text_embed = TextEmbedding(dim, text_num_embeds) |
|
self.audio_embed = AudioEmbedding(mel_dim, dim) |
|
|
|
self.rotary_embed = RotaryEmbedding(dim_head) |
|
|
|
self.dim = dim |
|
self.depth = depth |
|
|
|
self.transformer_blocks = nn.ModuleList( |
|
[ |
|
MMDiTBlock( |
|
dim = dim, |
|
heads = heads, |
|
dim_head = dim_head, |
|
dropout = dropout, |
|
ff_mult = ff_mult, |
|
context_pre_only = i == depth - 1, |
|
) |
|
for i in range(depth) |
|
] |
|
) |
|
self.norm_out = AdaLayerNormZero_Final(dim) |
|
self.proj_out = nn.Linear(dim, mel_dim) |
|
|
|
def forward( |
|
self, |
|
x: float['b n d'], |
|
cond: float['b n d'], |
|
text: int['b nt'], |
|
time: float['b'] | float[''], |
|
drop_audio_cond, |
|
drop_text, |
|
mask: bool['b n'] | None = None, |
|
): |
|
batch = x.shape[0] |
|
if time.ndim == 0: |
|
time = repeat(time, ' -> b', b = batch) |
|
|
|
|
|
t = self.time_embed(time) |
|
c = self.text_embed(text, drop_text = drop_text) |
|
x = self.audio_embed(x, cond, drop_audio_cond = drop_audio_cond) |
|
|
|
seq_len = x.shape[1] |
|
text_len = text.shape[1] |
|
rope_audio = self.rotary_embed.forward_from_seq_len(seq_len) |
|
rope_text = self.rotary_embed.forward_from_seq_len(text_len) |
|
|
|
for block in self.transformer_blocks: |
|
c, x = block(x, c, t, mask = mask, rope = rope_audio, c_rope = rope_text) |
|
|
|
x = self.norm_out(x, t) |
|
output = self.proj_out(x) |
|
|
|
return output |
|
|